Vassilios V. Dimakopoulos

On the Performance of Flooding-Based Resource Discovery

We consider flooding-based resource discovery in distributed systems. With flooding, a node searching for a resource contacts its neighbors in the network, which in turn contact their own neighbors and so on until a node possessing the requested resource is located. Flooding assumes no knowledge about the network topology or the resource distribution thus offering an attractive means for resource discovery in dynamically evolving networks such as peer-to-peer systems. We provide analytical results for the performance of a number of flooding-based approaches that differ in the set of neighbors contacted at each step. The performance metrics we are interested in are the probability of locating a resource and the average number of steps and messages for doing so. We study both uniformly random resource requests and requests in the presence of popular (hot) resources. Our analysis is also extended to take into account the fact that nodes may become unavailable either due to failures or voluntary departures from the system. Our analytical results are validated through simulation

Creating and maintaining replicas in unstructured peer-to-peer systems

In peer-to-peer systems, replication is an important issue as it improves search performance and data availability. It has been shown that optimal replication is attained when the number of replicas per item is proportional to the square root of their popularity. In this paper, we focus on updates in the case of optimal replication. In particular, we propose a new practical strategy for achieving square root replication called pull-then-push replication (PtP). With PtP, after a successful search, the requesting node enters a replicate-push phase where it transmits copies of the item to its neighbors. We show that updating the replicas can be significantly improved through an update-push phase where the node that created the copies propagates any updates it has received using similar parameters as in replicate-push. Our experimental results show that replicate-push coupled with an update-push strategy achieves good replica placement and consistency with small message overhead.

A Peer-to-Peer Approach to Resource Discovery in Multi-agent Systems

A multi-agent system is a network of software agents that cooperate to solve problems. In open multi-agent systems, the agents that need resources provided by other agents are not aware of which agents provide the particular resources. We propose a fully distributed approach to this resource discovery problem. Each agent A maintains a limited size local cache in which it keeps information about k different resources, that is, for each of the k resources, it stores the contact information of one agent that provides it. The agents in the cache of agent A are called As neighbors. An agent searching for a resource contacts its local cache and if there is no information for the resource, it contacts its neighbors, which in turn contact their neighbors and so on until the resource is found in some cache. We consider variations of this flooding-based search and develop and verify by simulation analytical models of their performance for both uniformly random resource requests and for requests in the case of hot spots. Finally, we introduce two approaches to the problem of updating the caches: one that uses flooding to propagate the updates and one that builds on the notion of an inverted cache.

A microbenchmark study of OpenMP overheads under nested parallelism

In this work we present a microbenchmark methodology forassessing the overheads associated with nested parallelism in OpenMP.Our techniques are based on extensions to the well known EPCC microbenchmarksuite that allow measuring the overheads of OpenMPconstructs when they are effected in inner levels of parallelism. Themethodology is simple but powerful enough and has enabled us to gaininteresting insight into problems related to implementing and supportingnested parallelism. We measure and compare a number of commercialand freeware compilation systems. Our general conclusion is that whilenested parallelism is fortunately supported by many current implementations,the performance of this support is rather problematic. Thereseem to exist issues which have not yet been addressed effectively, asmost OpenMP systems do not exhibit a graceful reaction when made toexecute inner levels of concurrency.

Nested parallelism in the OMPI OpenmP/C compiler

This paper presents a new version of the OMPi OpenMP C compiler, enhanced by lightweight runtime support based on user-level multithreading. A large number of threads can be spawned for a parallel region and multiple levels of parallelism are supported efficiently, without introducing additional overheads to the OpenMP library. Management of nested parallelism is based on an adaptive distribution scheme with hierarchical work stealing that not only favors computation and data locality but also maps directly to recent architectural developments in shared memory multiprocessors. A comparative performance evaluation of several OpenMP implementations demonstrates the efficiency of our approach.

A theory for total exchange in multidimensional interconnection networks

Total exchange (or multiscattering) is one of the important collective communication problems in multiprocessor interconnection networks. It involves the dissemination of distinct messages from every node to every other node. We present a novel theory for solving the problem in any multidimensional (cartesian product) network. These networks have been adopted as cost-effective interconnection structures for distributed-memory multiprocessors. We construct a general algorithm for single-port networks and provide conditions under which it behaves optimally. It is seen that many of the popular topologies, including hypercubes, k-ary n-cubes, and general tori satisfy these conditions. The algorithm is also extended to homogeneous networks with 2/sup k/ dimensions and with multiport capabilities. Optimality conditions are also given for this model. In the course of our analysis, we also derive a formula for the average distance of nodes in multidimensional networks; it can be used to obtain almost closed-form results for many interesting networks.

Performance analysis of distributed search in open agent systems

In open multi-agent systems agents need resources provided by other agents but they are not aware of which agents provide the particular resources. Most solutions to this problem are based on a central directory that maintains a mapping between agents and resources. However, such solutions do not scale well since the central directory becomes a bottleneck in terms of both performance and reliability. In this paper, we introduce a different approach: each agent maintains a limited size local cache in which it keeps information about k different resources, that is, for each of k resources, it stores the contact information of one agent that provides it. This creates a directed network of caches. We address the following fundamental problem: how can an agent that needs a particular resource find an agent that provides it by navigating through this network of caches? We propose and analytically compare the performance of three different algorithms for this problem, flooding, teeming and random paths, in terms of three performance measures: the probability to locate the resource, the number of steps and the number of messages to do so. Our analysis is also applicable to distributed search in unstructured peer-to-peer networks.

Cache updates in a peer-to-peer network of mobile agents

In open multi-agent systems, agents need resources provided by other agents but they are not aware of which agents provide particular resources. We consider a peer-to-peer approach, in which each agent maintains a local cache with information about k resources, that is for each of the k resources, an agent that provides it. However, when an agent or a resource moves, cache entries become obsolete. We propose a suite of cache update policies that combine pull-based invalidation that is initiated by the agent that maintains the cache with push-based invalidation that is initiated by the agent that moves. We study and compare variations of oblivious flooding-based push/pull along with an informed push approach where each agent maintains a list of the agents that have it cached. Our experimental results indicate that a novel variation of flooding for push where a moving agent propagates its new location to agents in its old neighborhood achieves good cache consistency with a small message overhead. The proposed policies are suitable for any peer-to-peer system where peers cache information about other peers and this information becomes obsolete.

The SMart autonomous storage (SmAS) system

The increasing demand for storage capacity and throughput has generated a need for storage architectures that scale their processing power with the growing size of datasets. In this paper, we give an overview of the SmAS system that employs network attached disks with processing capabilities. In the SmAS system, users can deploy and execute code at the disk. At-the-disk executed application code is written in a stream-based language that enforces code security and bounds the codes memory requirements. The SmAS disks system software provides basic support for process scheduling and memory management. We present an initial implementation of the system and report performance results that validate our approach.

On TSC checkers for m-out-of-n codes

Paschalis et al. (1988) have given a structured method to design TSC m-out-of-2m code checkers suitable for VLSI implementation. In this correspondence we give sufficient conditions so that the method previously given can be used to design checkers for classes of m-out-of-n codes with n/spl ne/2m. >